Individual religious affiliation, religious community context, and reproductive health care in Mozambique*

Individual religious affiliation, religious community context, and reproductive health care in Mozambique* Boaventura Cau** Victor Agadjanian Center for Population Dynamics Arizona State University * The support of the Eunice Kennedy Shriver National Institute of Child Health & Human Development (grant # R01 HD050175) is gratefully acknowledged. **Direct correspondence to Boaventura Cau, Center for Population Dynamics, School of Social and Family Dynamics, Arizona State University, Tempe, Arizona, 85287-3701; email:boaventura.cau@asu.edu

Individual religious affiliation, religious community context, and reproductive health care in Mozambique Abstract Most studies of religion and demographic behavior and outcomes in sub-saharan Africa focus on effects of individual religious affiliation. In the present study we use recent data from a survey of over 2000 women and a survey of over 1000 religious congregations from southern Mozambique to examine the effects of women s religious affiliation and religious community environment on two indicators of reproductive health care utilization: child delivery in a health care facility and the number of prenatal consultations. Our findings suggest a beneficial effect of affiliation with Catholic and Mission Protestant denominations for child delivery in a health facility and a favorable effect of affiliation with any organized religion for prenatal visits. We also find a significant positive association between the presence of Catholic and Mission Protestant denominations within community with child delivery in a health care unit, regardless of individual religious affiliation, but no clear pattern for the effect of religious community context on prenatal consultations. 2

Theoretical background In sub-saharan Africa most people are affiliated with organized religion and view religion as having guiding power in their lives (e.g., Carmody, 2003; Gallup-International, 2010; Garner, 2000; Gyimah, Takyi and Addai, 2006; Trinitapoli, 2006). Researchers have sought to understand the relationship between religion and various demographic aspects such as HIV/AIDS (e.g., Agadjanian, 2005; Agadjanian and Menjívar, 2008; Lagarde et al., 2000; Maman et al., 2009; Rankin et al., 2005; Takyi, 2003; Trinitapoli, 2009; Trinitapoli and Regnerus, 2006), child survival (e.g., Antai et al., 2009; Cau, Sevoyan and Agadjanian, 2010; Gregson et al., 1999; Gyimah, 2007), fertility and contraceptive use (e.g., Adongo, Phillips and Binka, 1998; Agadjanian, 2001; Agadjanian, Yabiku and Fawcett, 2009; Avong, 2001; Yeatman and Trinitapoli, 2008) and health care services utilization (Gyimah, Takyi and Addai, 2006). Typically, studies entertain three alternative hypotheses for the association of religion and demographic behavior and outcomes: 1) the particularistic theology hypothesis, which relates demographic behavior and outcomes to explicit religious prescriptions and proscriptions; 2) the minority-group status hypothesis, which focuses on a religious group socioeconomic and/or political disadvantages; and 3) the characteristics hypothesis, which posits that demographic differences among religious groups can be fully explained by other factors (Goldscheider, 1971). However, most studies, regardless of their theoretical perspective, have focused on individual religious affiliation without taking into consideration how the community religious context may influence demographic outcomes of both affiliated and unaffiliated individuals. Specifically, it has not been sufficiently investigated whether the presence and configuration of religious congregations in a community may have effects on health regardless of individual affiliation. In this study, we use recent unique individual- and congregation-level data collected in Mozambique, to examine individual and contextual effects of religion on reproductive health care utilization. 1

Previous literature that has looked at the association between religion and health-related outcomes in sub-saharan Africa has painted a mixed picture. On the association between religion and HIV/AIDS for example, Lagarde and colleagues examined the role of religion on preventative behavior against HIV/AIDS in Senegal, a country predominantly Muslim but with a Christian population; they found religion to be negatively associated with preventative behaviors (Lagarde et al., 2000). Takyi (2003) investigated the association between women s religious affiliation and their knowledge of AIDS and practice of preventive behaviors against AIDS in Ghana to find that Christian women were more likely to report more knowledge of HIV than non- Christians. But there were inconsistent findings pertaining to the effect of religious affiliation on condom use and decrease in sexual partners (Takyi, 2003). In respect to child health, Gyimah found some association of religious affiliation with child survival in Ghana but that association disappeared after controlling for socioeconomic factors, especially education (Gyimah, 2007). Several studies have argued that the effect of religion on health outcomes, such as child survival, operates through differential access to health services. Thus, in Nigeria, Antai and colleagues looked at the effect of mother s religion affiliation on under-five mortality and concluding that an association of belonging to a traditional religion and mortality was explained by religion-based differential use of maternal and child health services (Antai et al., 2009). In Ghana, Gyimah and colleagues showed significantly higher utilization of health services among Catholic mothers compared to unaffiliated mothers (Gyimah, Takyi and Addai, 2006). While these investigations further our understanding of the relevance of religious affiliation on some health outcomes in sub-saharan Africa, we need to better understand religious community context and its possible effect on health-related outcomes. In this study, in seeking to understand the role of community religious milieu in reproductive health care utilization, we are informed by the literature on contextual and social influences on fertility and contraceptive behavior and HIV/AIDS-related outcomes (Agadjanian, Yabiku and Fawcett, 2009; Helleringer & Kohler, 2005; Kohler, 1997; Kohler, Behrman & Watkins, 2007; Smith & 2

Watkins, 2005), and broader analyses of diffusion of innovations and demographic change (Rogers, 1995). Research has identified woman s formal education as one of drivers of reproductive health-care services utilization (Elo, 1992; Fosu, 1994; Onah, Ikeako and Iloabachie, 2006; Caldwell, 1979). Studies in sub-saharan Africa have also observed variations among religious congregations in average level of schooling of their members, with Mainline churches (Catholic and Mission Protestant churches) typically having better educated members (Agadjanian, 2001; Gymah, Takyi and Addai, 2006; Kollehlon, 1994; Takyi, 2003; Zou et al., 2009). However, it is likely that women s health-related knowledge is acquired not only through formal education but also through informal learning that occurs when mothers interact with each other in diverse situations. For example, in a religious environment of developing settings, relatively lower educated mothers may interact with relatively highly educated mothers (including nurses, school teachers), which may lead to informal exchanges of health-related information and experiences. Given the importance of religion and religious participation in sub-saharan women s lives, much of these exchanges may occur in religion-related networks. As members of diverse religious organizations often live in the same community, informal social interaction among them might occur frequently. Some studies suggest that social interactions might lead to diffusion of health ideas within communities. Using a household survey of about 2500 residents of coastal Ghana, Andrzejewski, Reed and White (2009) examined the effect of community context on adults knowledge of etiology and prevention of child illnesses and they reported a positive association between the proportion of literate individuals in the community and health knowledge. These authors argued that living in a community with high levels of literacy increased health knowledge even for illiterate community members. In rural Guatemala, Goldman, Pebley and Beckett (2001) investigated whether women who had social contacts outside of the community were more likely to hold the belief in inadequate hygiene as cause of diarrheal illness and to use 3

hygienic practices. The researchers reported that having a relative living abroad or in the city of Guatemala was significantly and positively associated to holding beliefs related to hygiene or contamination and also to practicing hygienic behavior. Additionally, the authors found that participation in community groups (e.g., holding a directive position in a religious organization) was significantly associated with possessing beliefs related to hygiene or contamination but this finding was not consistent across the health measures. Agadjanian and Menjívar (2008) in their study in Mozambique showed how church-based informal social interactions help women in coping with their worries and misgivings about HIV/AIDS. In this study, we contribute to the literature on the relationship between religion and health in sub-saharan Africa by examining how women s religious affiliation affects their use of prenatal and child delivery services and whether the community religious environment, approximated by the spatial configuration of religious congregations, influences use of these services. Hypotheses Informed by the literature reviewed above, we formulate two main hypotheses. First, we expect that women belonging to a religious congregation will have a higher likelihood of delivering in health facilities and will have a higher number of antenatal consultations than women who are not affiliated with organized religion. As part of the first hypothesis, we also expect that Catholics and Mainline Protestants will display a particular advantage. This expectation is based on the literature reviewed above as well as our earlier investigation of denominational differences in under-five mortality (Cau, Sevoyan and Agadjanian, 2010). Our second hypothesis is that the higher the number of religious congregations in a community, the greater the probability that women living in that community, whatever their religious affiliation, will give birth in a clinic, and that this number will be positively associated with the number of prenatal consultations, regardless of other factors. Again, as a sub- 4

hypothesis, we posit that the presence of congregations of Mainline denominations (Catholic and Mission Protestant churches) will be particularly beneficial given their greater socioeconomic diversity and a relatively high proportion of educated members (e.g., nurses and teachers.). The presence of Mainline churches may facilitate the spread of attitudes and preferences that encourage reproductive health care utilization among the members of those churches and non-members alike. Although we cannot observe the process of religion-based information exchange and influence directly with our data, we assume based on the literature that this process would be at the root of whatever individual and community level differentials that we might detect after controlling for standard sociodemographic factors. Setting Mozambique is a nation in Southern Africa with 21.9 million inhabitants in 2009 (INE, 2010). It became independent from Portugal in 1975 and then went through a civil war between 1976 and 1992 that destroyed much of the social and economic infrastructure of the country (Abrahamsson and Nilsson 1995; Minter 1994). The post-civil war reconstruction in Mozambique has been remarkable, with economic growth averaging 8% per year between 1997 and 2003. However, the country remains one of poorest in the world, with the Gross National Income per capita of $380 in 2008, life expectancy at birth of 48 years in 2008 (World Bank, 2010). The setting of this study is Chibuto district of Gaza province in Southern Mozambique (Figure 1). Ethnically homogenous, Chibuto district had a population of about 165,000 in 2007 (INE, 2009), one third of it living in the only urban area of the district (its headquarters). The economy of Chibuto is based on subsistence agriculture and labor migration remittances, mainly from South Africa. The religious landscape of Chibuto district is dominated by Catholics, Mainline or Mission Protestants (e.g., Presbyterian, Anglican, Methodist and Baptist) and a growing number of small Pentecostal-type churches mainly Zionist churches, Apostolic 5

churches and Assemblies of God (Agadjanian, 2005; Agadjanian and Menjivar, 2008). Zionist churches are a growing group of small Pentecostal churches originating in southern Africa that emphasize miracle healing through prayer and Holy Spirit (Agadjanian, 2001; Gregson et al., 1999; Turner, 1967). [Figure 1 about here] Data and Methods This study uses data from an individual survey and an institutional survey conducted in 2008 in Chibuto district. The individual survey was a representative-based cluster survey of 2019 women aged 18-50, both affiliated and nonaffiliated with a religious denomination. In 82 randomly selected communities (clusters) located in both urban and rural areas of the district, the individual survey collected information on respondents religious affiliation history since birth until the year of the survey. The institutional survey conducted in parallel with the individual survey included all the congregations active in the district more than a thousand different congregations in total. The survey collected a variety of information about the congregations, including their geographic coordinates. The study also uses geographic coordinates of health units in Chibuto district and three other districts that share border with Chibuto (districts of Chókwè, Guijá and Manjacaze). Measures Outcome variables Two measures of reproductive health care utilization constitute our outcome variables: i) whether a woman s youngest living child was born in a health care facility and ii) the number of prenatal consultations a woman attended before the birth of her youngest child. Mozambique, like much of sub-saharan Africa has very high maternal mortality and morbidity (Hill et al., 2007; Gage, 2007; WHO, 2004), and it is assumed that use of reproductive health care services may 6

improve the health of women and their children (Short and Zhang, 2004). To be able to use the information on congregations collected through the institutional survey and the current characteristics from the individual surveys and to minimize problems associated with recall, the study looks only at women whose youngest children were born in 5 years before the survey. Predictor variables We use two predictors individual religious affiliation and religious community context. Individual affiliation includes the following categories: 1. Catholic or Mainline Protestant; 2. Zionist; 3. Other Pentecostal or Similar; and 4. Non-affiliated. Affiliation is lagged one year before the birth of the youngest child. It is based on a question about religious life history of the respondent. The religious community context measure is build from the geographic coordinates of congregations and of respondents households. Using the Geodist function in SAS 9.2, distance from each respondent s household in the sample to each congregation in each category of religious congregation (Mainline churches, Zionist churches and Other Pentecostal churches) was calculated. Following this, the congregation s concentration measure was created based on the number of congregations of a given type that exist with 5 Km from each respondent s household for rural areas and 1 Km in urban area. We assumed that within 5 Km of distance people in rural areas are able to interact with each other (for example, people may frequently walk within 5 Km to the field or to church). Because of relatively a high concentration of people in urban areas, interaction among individuals within 1 Km of distance from each other was considered to be equivalent to that which occurs within 5 Km in rural areas. As individual controls we use woman s age, marital status, education, and the number of children. We also include location of respondent s household relative to the closest clinic, which is operationalized as a set of dummy variables: urban (no or negligible distance); rural less than 5 Km, and rural 5 Km or more). We also control for average literacy level in each community (individual survey cluster). For analysis we use SAS 9.2 Glimmix procedure to fit a random 7

intercept logistic regression for the outcome whether the respondent s last child was born in a health care establishment and random intercept ordinary linear regression for the outcome about the number of prenatal consultations that the respondent attendant before the birth of her last child. A model with random intercept is employed because respondents are clustered within communities and therefore may share some unobserved characteristics, which confound the results. Preliminary results Descriptive results Table 1 and 2 show key descriptive statistics characterizing the various groups in our sample. There is variation in proportion of children born in health facility between denominational groups. Among members of Mainline churches, about 74% had their youngest child born in a health care establishment. Following Mainline church goers, members belonging to Other Pentecostal churches had about 65.12%. The category of No Affiliation exhibits the smallest percentage of members who had their last child born in a health facility (about 49%). Similar pattern is observed for the mean number of prenatal consultations that a mother attended before the birth of her last child: women belonging to Mainline churches show the highest mean number of prenatal consultations (4.44) and nonaffiliated women the lowest (3.60). The mean age of respondents in the sample does not vary across religious denomination groups, ranging from 30 years in the Mainline religious group to 32 years in the No Affiliation group. All religious groups have slightly similar number of living children (approximately 3 children). Our sample is dominated by married women, with at least one forth of women in each denomination group married. Differences in educational achievement of members between groups are notable. This suggests variation in socioeconomic background among members belonging to distinct groups. 8

The religious group of No affiliation is overwhelmingly composed by illiterate members (about 56%). It is followed by the Zionist group in terms of presence of members without any schooling. The average level of education among communities (sample clusters) of the study area is 3 years. Table 2 displays the mean number of religious congregations in each denominational category within 5 Km in rural areas (1 Km in the urban area) from respondents residence within the district boundaries. Other Pentecostal churches are the most prevalent congregations in both rural and urban areas (5.3 churches per 5 Km and 6.3 churches per 1 Km, respectively). Zionist congregations follow Other Pentecostal churches and are more common in urban areas (with 5.21 churches per 1 Km). For Mainline churches, on average there are approximately three churches within the defined distance from respondents households. [Table 1 and 2 about here] Multivariate results Whether the youngest child was born in a health care facility Table 3 shows results from random intercept logistic regression models examining the association between the religious affiliation of the mother, denominational concentration in the community and whether or not the youngest child was delivered in a health care establishment. Model 1 assesses the effect of maternal religious affiliation and it shows that overall religion appears to have a favorable effect on the odds of child delivery in a health facility, but it is only significant for mothers affiliated to Mainline denominations. Compared to non-affiliated women (the reference category), children of women affiliated with Mainline churches are 2 times likely to have been delivered in a health facility. [Table 3 about here] 9

In model 2, we try to capture the independent effect of maternal religion on youngest child delivery in a health care establishment controlling for socioeconomic covariates. The effect of mother s religious affiliation decreases in magnitude for all denominations but remains significant for Mainline churches. While affiliation with a Zionist and Other Pentecostal denominations appears to increase the odds that the youngest child was delivered in a health facility, mothers affiliated to these congregations are not significantly different from unaffiliated mothers. These findings seem to support the hypothesis that religion has its own effect on child delivery in a health facility, particularly for mothers affiliated to Mainline churches. Also notable in model 2 is that education has a significant positive effect on the odds of the youngest child having been delivered at a health care establishment. Thus, children of mothers with 5 and more years of schooling are more than 2.4 times likely to have been born in a health facility than children of women with no education. In model 3 we attempt to assess whether living in a community with a higher number of religious organizations has an effect on child delivery in a health facility. No other variables are included. The findings show a positive and significant effect of the presence of Mainline denominations. On average, each additional Mainline church in the community increases the odds that the youngest child was born in a health facility by 14% (OR=1.14). When we added maternal religious affiliation and other characteristics and a community-level control in model 4, the coefficient for the prevalence of Mainline churches retained its magnitude and significance. One interesting finding from model 4 is that the prevalence of Zionist congregations significantly decreases the odds that the youngest child was delivered in a health facility. To be sure that these effects do not simply reflect the fact that Mainline and Zionist churches are disproportionately concentrated in socioeconomically more and less advanced areas, respectively, we control for the average educational level in the community. Notable in model 4 is that individual religion is still positively associated with the odds of giving birth in a health clinic; but it is only significant for mothers belonging to Mainline churches 10

(p <.1). The findings from model 4 partially support our second hypothesis that higher number of religious congregations in the community increases the odds of giving birth in clinic: while the concentration of Mainline congregations displays a positive association, that of Zionist congregations shows a negative effect. Our sub-hypothesis for the second hypothesis positing a particular large advantage of Catholic and Mission Protestant churches for delivering in clinics therefore finds support. Number of prenatal consultations before the birth of the youngest child In Table 4 we assess the association between the religious affiliation of the mother, religious community environment and the number of prenatal consultations attended before the birth of the youngest child. The first model in Table 4 includes maternal religious affiliation as the only predictor. In the second model socioeconomic covariates are added as controls. The results in model 1 show that religion has a positive and significant effect on the use of prenatal services. Compared to unaffiliated mothers, mothers identifying themselves with any denomination were significantly likely to have attended more prenatal visits before the birth of the youngest child. The positive effect of religious affiliation on the use of prenatal services remains unchanged in model 2. Compared to unaffiliated women, women belonging to any denomination are more likely to have attended higher number of prenatal consultations. Coefficients for all the categories of mother s religious affiliation decrease in magnitude and for Zionist affiliation it becomes marginally significant. However, no advantage of Mainline church members of Other Pentecostals and Zionist is noticeable. In Model 3 and 4 we examine the contribution of community religious context to predicting the number of women s prenatal consultations. Although the results are not statistically significant, Model 1 suggests that the concentration of Mainline churches and Other Pentecostal churches in the community has positive effect and the predominance of Zionist churches a negative effect on prenatal visits. This pattern holds after controlling for women s 11

individual characteristics and community average level of literacy, and the presence of Zionist churches reaches a marginal significance (model 4). An interesting observation in model 4 is that women affiliated to any religious denomination continue to have significant advantage over unaffiliated women. Findings in table 4 support the hypothesis that in this setting maternal religion has a positive effect on number of prenatal visits net of socioeconomic characteristics. The effect of religious community context on prenatal visits is unclear; while the direction of the coefficients in model 4 points to the expectations for the predominance of Mainline and Other Pentecostal denominations, for the prevalence of Zionist churches the effect is in opposite direction of the one hypothesized. [Table 4 about here] Next steps To prepare the paper for presentation at the PAA Meeting we plan to fine-tune the specification of our models and expand our analysis. Specifically we will develop and test alternative definitions of community religious milieu. Also, because the location of churches and maternal and child clinics in the study area may not be random, in our final paper we will control for distance between clinics and churches. 12

References Abrahamsson, H. and Nilsson, A. (1995). Mozambique: the troubled transition: from socialist construction to free market capitalism. London: Zed Books. Adongo, P.B.; Phillips, J.F. and Binka, F.N. (1998). The influence of traditional religion on fertility among the Kassena-Nankana of Northern Ghana. Studies in Family Planning, 29(1):23-40. Agadjanian, V. (2001). Religion, social milieu, and the contraceptive revolution. Population Studies, 55:135-148. Agadjanian, V. (2005). Gender, religious involvement, and HIV/AIDS prevention in Mozambique. Social Science & Medicine, 61:1529-1539. Agadjanian, V. and Menjívar, C. (2008). Talking about the Epidemic of the Millennium : religion, informal communication, and HIV/AIDS in sub-saharan Africa. Social Problems, 55(3):301-321. Agadjanian, V.; Yabiku, S. and Fawcett, L. (2009). History, community milieu, and Christian- Muslim differentials in contraceptive use in sub-saharan Africa. Journal for the Scientific Study of Religion, 48(3): 462-79. Andrzejewski, C.S.; Reed, H.E. and White, M.J. (2009). Does where you live influence what you know? Community effects on health knowledge in Ghana. Health and Place, 15:228-238. Antai, D.; Ghilagaber, G.; Wedrén, S.; and Macassa, G. (2009). Inequalities in under-five mortality in Nigeria: differentials by religious affiliation of the mother. Journal of Religion and Health, 48(3):290-304. Avong, H.N. (2001) Religion and fertility among the Atyap in Nigeria. Journal of Biosocial Sciences, 33:1-12. Caldwell, J. (1979). Education as a factor in mortality decline. An examination of Negerian data. Population Studies, 33(3):395-413. Carmody, B. (2003). Religious education and pluralism in Zambia. Religious Education, 98(2):139-154. 13

Cau, B.; Sevoyan, A. and Agadjanian, V. (2010). Religion, child mortality and health in Mozambique. Presented at the Population Association of America 2010 Annual Meeting, Dallas, Texas April 15-17, 2010. Elo, I.T. (1992). Utilization of maternal health-care services in Peru: the role of women s education. Health Transition Review, 2(1): 1-20. Fosu, G.B. (1994). Childhood mortality and health services utilization: cross-national comparisons of user-related factors from DHS data. Social Science and Medicine, 38(9): 1209-1220. Gage, A.J. (2007). Barriers to the utilization of maternal health care in rural Mali. Social Science and Medicine, 65:1666-1682. Gallup-International (2010). Religion in the world at the end of the millennium. Gallup International. Available online at http://www.gallup-international.com/. Retrieved in September, 2010. Garner, R.C. (2000). Religion as a source of social change in new South Africa. Journal of Religion in Africa, 30(3):310-343. Goldman, N.; Pebley, A.R. and Beckett, M. (2001). Diffusion of ideas about personal hygiene and contamination in poor countries: evidence from Guatemala. Social Science and Medicine, 52:53-69. Goldscheider, C. (1971). Population, modernization, and social structure. Boston: Little, Brown and Company. Gregson, S.; Zhumu, T.; Anderson, R.M. and Chandiwana, S.K. (1999). Apostles and Zionists: the influence of religion on demographic change in Zimbabwe. Population Studies, 53:179-193. Gyimah, S.O. (2007). What has faith got to do with it? Religion and child survival in Ghana. Journal of Biosocial Science, 39:923-937. 14

Gyimah, S.O.; Takyi, B.K. and Addai, I. (2006). Challenges to the reproductive-health needs of African women: on religion and maternal health utilization in Ghana. Social Science and Medicine, 62:2930-2944. Helleringer, S. and Kohler, H. (2005). Social networks, perceptions of risk, and changing attitudes towards HIV/AIDS: new evidence from a longitudinal study using fixed-effects analysis. Population Studies, 59(3):265-282. Hill, K.; Thomas, K.; AbouZahr, C.; Walker, N.; Say, L.; Inoue, M.; Suzuki, E. (2007). Estimates of maternal mortality worldwide between 1990 and 2005: an assessment of available data. Lancet, 370:1311-19. INE (Instituto Nacional de Estatística) (2009). Snopse dos resultados definitivos do terceiro censo geral de população. Maputo: Instituto Nacional de Estatística. INE (Instituto Nacional de Estatística) (2010). Estatísticas de Moçambique. Instituto Nacional de Estatística. Available online at http://www.ine.gov.mz/. Retrieved in September 2010. Kohler, H. (1997). Learning in social networks and contraceptive choice. Demography, 34(3):369-383. Kohler, H., Behrman, J. & Watkins, S.C. (2007). Social networks and HIV/AIDS risk perceptions. Demography, 44(1): 1-33. Kollehlon, K.T. (1994). Religious affiliation and fertility in Liberia. Journal of Biosocial Sciences, 26:493-507. Lagarde, E.; Enel, C.; Seck, K.; Gueye-Ndiaye, A.; Piau, J.; Pison, G.; Delaunay, V. et al. (2000). Religion and protective behaviours towards AIDS in rural Senegal. AIDS, 14(13):2027-2033. Maman, S.; Cathcart, R.; Burkhardt, G.; Omba, S. and Behets, F. (2009). The role of religion in HIV-positive women disclosure experiences and coping strategies in Kinshasa, Democratic Republic of Congo. Social Science and Medicine, 68:965-970. 15

Minter, W. (1994). Apartheid s Contras: an inquiry into the roots of war in Angola and Mozambique. London: Zed Books. Onah, H.E.; Ikeako, L.C. and Iloabachie, G.C. (2006). Factors associated with the use of maternal services in Enugu, southeast Nigeria. Social Science and Medicine, 63:1870-1878. Rankin, S.; Lindgren, T.; Rankin, W.W. and Ng oma, J. (2005). Donkey work: women, religion, and HIV/AIDS in Malawi. Health Care for Women International, 26:4-16. Rogers, E.M. (1995). Diffusion of innovations. New York: Free Press. Short, S.E. and Zhang, F. (2004). Use of maternal health services in China. Population Studies, 58(1):3-19. Smith, K.P. & Watkins, S.C. (2005). Perceptions of risk and strategies for prevention: responses to HIV/AIDS in rural Malawi. Social Science & Medicine, 60: 649-660. Takyi, B.K. (2003). Religion and women s health in Ghana: insights into HIV/AIDS preventive and protective behavior. Social Science and Medicine, 56:1221-1234. The World Bank (2010). World Development Indicators, Mozambique. Available online at http://data.worldbank.org/country/mozambique. Retrieved in September 2010. Trinitapoli, J. (2006). Religious responses to AIDS in sub-saharan Africa: an examination of religious congregations in rural Malawi. Review of Religious Research, 47(3):253-270. Trinitapoli, J. (2009). Religious teachings and influences on the ABCs of HIV prevention in Malawi. Social Science and Medicine, 69:199-209. Trinitapoli, J. and Regnerus, M.D. (2006). Religion and HIV risk behaviors among married men: initial results from a study in rural sub-saharan Africa. Journal for the Scientific Study of Religion, 45(4):505-528. Turner, H.W. (1967). A typology for African religious movements. Journal of Religion in Africa, 1(1):1-34. WHO [World Health Organization] (2004). Maternal mortality in 2000: estimates developed by WHO, UNICEF and UNFPA. World Health Organization. 16

Yeatman, S.E. and Trinitapoli, J. (2008). Beyond denomination: the relationship between religion and family planning in rural Malawi. Demographic Research, 19(55):18-51-1882. Zou, J.; Yamanaka, I.; John, M.; Watt, M.; Ostermann, J. and Thielman, N. (2009). Religion and HIV in Tanzania: influence of religious beliefs on HIV stigma, disclosure, and treatment attitudes. BMC Public Health 2009, 9(75):1-12. 17

Figure 1. Map of Chibuto district 18

Table 1. Selected descriptive statistics (percentage, unless stated otherwise) Woman's religious affiliation Variable Mainline Zionist Other Pentecostal No Affiliation Last child born in health clinic 74.07 58.94 65.12 49.47 Woman's number of prenatal consultations (mean) 4.44 3.98 4.30 3.60 Woman's age (mean) 30.41 31.18 31.18 32.41 Woman's number of children (mean) 2.71 2.75 2.89 2.89 Married woman 76.07 76.12 78.13 81.97 Women's education None 16.41 38.37 26.69 56.12 1 to 4 years 36.62 39.08 38.05 32.07 5+ years 46.97 22.55 35.26 11.81 Percentage of denominational group in the sample 19.66 41.90 25.36 11.54 Table 2. Concentration of religious congregations by denomination (means) Place of residence Mainline Zionist Other Pentecostal Rural 2.94 2.98 5.30 Urban 1.69 5.21 6.26 Note: Within 5 Km in rural areas and 1 Km in urban areas from respondents residence within District boundaries 19

Table 3. Random intercept logistic regression predicting youngest child delivery in a health care facility with mother's religious affiliation and concentration of churches as main predictors (odds ratio) Predictors and controls Model 1 Model 2 Model 3 Model 4 Mother's religious affiliation No Affiliation [reference] 1 1 1 Mainline churches 2.00 ** 1.57 * 1.47 Zionist churches 1.17 1.08 1.06 Other Pentecostal churches 1.22 1.06 1.04 Community concentration of religious congregations Mainline churches 1.14 ** 1.14 ** Zionist churches 1.00 0.94 * Other Pentecostal churches 1.02 1.06 Individual (mother) controls Mother's age 0.99 0.99 Mother s number of children 0.89 * 0.89 * Mother's marital status Not married [reference] 1 1 Married 1.14 1.16 Mother's education Less than 1 [reference] 1 1 1-4 years education 1.54 ** 1.54 ** 5+ years education 2.42 ** 2.19 ** Distance to closest clinic Urban [reference] 1 1 Rural less than 5 Km 1.12 0.90 Rural 5 Km or more 0.64 * 0.59 * Community context controls Average level of education in the village 1.18 Intercept 1.36 * 1.94 1.18 1.21-2 Res Log Pseudo-Likelihood 5693.69 5781.54 5708.36 5822.76 Generalized Chi-Square 1218.19 1223.25 1219.76 1224.63 Number of cases 1297 1297 1297 1297 Notes: - p<.1; *- p.05; **- p.01. 20

Table 4. Random intercept ordinary linear regression predicting the number of prenatal consultations before the birth of youngest child with mother's religious affiliation and concentration of churches as main predictors (coefficient and standard error in parentheses) Predictors Model 1 Model 2 Model 3 Model 4 Mother's religious affiliation No Affiliation [reference] 1 1 1 Mainline churches 0.41* (0.16) 0.33* (0.17) 0.28 (0.17) Zionist churches 0.27* (0.14) 0.25 (0.14) 0.24 (0.14) Other Pentecostal churches 0.37* (0.15) 0.34* (0.15) 0.31* (0.15) Community concentration of religious congregations Mainline churches 0.04 (0.03) 0.03 (0.03) Zionist churches -0.03 (0.02) -0.04 (0.02) Other Pentecostal churches 0.04 (0.04) 0.05 (0.03) Individual (mother) controls Mother's age 0.02 (0.01) 0.01 (0.01) Number of living children -0.06 (0.04) -0.06 (0.04) Mother's marital status Not married [reference] 1 1 Married -0.08 (0.14) -0.07 (0.14) Mother's education Less than 1 [reference] 1 1 1-4 years education 0.19 (0.13) 0.18 (0.13) 5+ years education 0.28* (0.15) 0.22 (0.15) Distance to the closest clinic Urban [reference] 1 1 Rural less than 5 Km 0.14 (0.20) 0.09 (0.22) Rural 5 Km or more -0.02 (0.16) 0.00 (0.20) Community context controls Average level of education in village 0.10 (0.07) Intercept 3.84** (0.12) 3.84** (0.33) 4.03** (0.10) 3.25** (0.41) -2 Res Log Pseudo- Likelihood 4762.19 4777.58 4777.27 4792.00 Generalized Chi-Square 3640.37 3614.43 3626.09 3594.2 Number of cases 1193 1193 1193 1193 Notes: - p<.1; *- p.05; **- p.01. 21